This paper presents an algorithm for viewpoint-based similarity discernment of linguistic concepts on Semantic Network Array Processor (SNAP). The viewpoint-based similarity discernment plays a key role in retrieving similar propositions. This is useful for advanced knowledge processing areas such as analogical reasoning and case-based reasoning. The algorithm assumes that a knowledge base is constructed for SNAP, based on information acquired from the WordNet linguistic database. The algorithm identifies paths on the knowledge base between each given concept and a given viewpoint concept, then computes a similarity degree between the two concepts based on the number of nodes shared by the paths. A small scale knowledge base was constructed and an experiment was conducted on a SNAP simulator that demonstrated the feasibility of this algorithm. Because of SNAP's scalability, the algorithm is expected to work similarly on a large scale knowledge base.

This paper presents a type inference algorithm for a logic language, HS. The algorithm uses a program transformation, SPS, to given programs as a type inference. This method is theoretically clear, because applying it to given programs is equal to executing it partially. No other additional framework is needed for our approach. In contrast, many studies on type inference for logic languages are based on Mycroft and O'Keefe's famous algorithm, which was initially developed for functional languages. Therefore, the meanings of the algorithms are theoretically unclear in the domain of logic languages. Our type inference is flexible. Users of the type inference system can specify the types of objects abstractly (weakly) if the types are not exactly known, or they can specify them particularly (strongly) if the types are exactly known. Both kinds of programs are inferred for types. In contrast, many type inference systems accept purely untyped programs. Thus, with these two features, our method is simple and flexible.

We discuss optimal rotation estimation from two sets of 3-D points in the presence of anisotropic and inhomogeneous noise. We first present a theoretical accuracy bound and then give a method that attains that bound, which can be viewed as describing the reliability of the solution. We also show that an efficient computational scheme can be obtained by using quaternions and applying renormalization. Using real stereo images for 3-D reconstruction, we demonstrate that our method is superior to the least-squares method and confirm the theoretical predictions of our theory by applying bootstrap procedure.

This article shows that a multivalued logic defined as juxtaposition of Boolean binary logics can use all of inference chain, induction and deduction that are important in realization of intelligent inference systems.

Access control is a key technology for providing data security in database management systems (DBMSs). Recently, various authorization models for object-oriented databases (OODBs) have been proposed since authorization models for relational databases are insufficient for OODBs because of the characteristics of OODBs, such as class hierarchies, inheritance, and encapsulation. Generally, an authorization is modeled as a set of rights, where a right consists of at least three components s, o, t and means that subject s is authorized to perform operation t on object o. In specifying authorizations implicitly, inference rules are useful for deriving rights along the class hierarchies on subjects, objects, and operations. An access request req=(s,o,t) is permitted if a right corresponding to req is given explicitly or implicitly. In this paper, we define an authorization model independent of any specific database schemas and authorization policies, and also define an authorization specification language which is powerful enough to specify authorization policies proposed in the literature. Furthermore, we propose an efficient access control method for an authorization specified by the proposed language, and evaluate the proposed method by simulation.

The adaptive associative memory proposed by Ma is used to construct a new model of semantic network, referred to as associative semantic memory (ASM). The main novelty is its computational effectiveness which is an important issue in knowledge representation; the ASM can do inference based on large conceptual hierarchies extremely fast-in time that does not increase with the size of conceptual hierarchies. This performance cannot be realized by any existing systems. In addition, ASM has a simple and easily understandable architecture and is flexible in the sense that modifying knowledge can easily be done using one-shot relearning and the generalization of knowledge is a basic system property. Theoretical analyses are given in general case to guarantee that ASM can flawlessly infer via pattern segmentation and recovery which are the two basic functions that the adaptive associative memory has.

We introduce a subclass of context-free languages, called pure context-free (PCF) languages, which is generated by context-free grammars with only one type of symbol (i. e. , terminals and nonterminals are not distinguished), and consider the problem of identifying paralleled even monogenic pure context-free (pem-PCF) languages, PCF languages with restricted and enhanced features, from positive data only. In this paper we show that the ploblem of identifying the class of pem-PCF languages is reduced to the ploblem of identifying the class of monogenic PCF (mono-PCF), by decomposing each string of pem-PCF languages. Then, with its result, we show that the class of pem-PCF languages is polynomial time identifiable in the limit from positive data. Further, we refer to properties of its identification algorithm.

We show that moving objects can be detected from optical flow without using any knowledge about the magnitude of the noise in the flow or any thresholds to be adjusted empirically. The underlying principle is viewing a particular interpretation about the flow as a geometric model and comparing the relative "goodness" of candidate models measured by the geometric AIC.

This paper proposes high-speed similitude retrieval schemes for a viewpoint-based similarity discrimination system (VB-SDS) and presents analytical and experimental performance evaluations. The VB-SDS, which contains a huge set of semantic definitions of commonly used words and computes semantic similarity between any two words under a certain viewpoint, promises to be a very important module in analogical and case-based reasoning systems that provide solutions under uncertainty. By computing and comparing similarities for all words contained in the system, the most similar word for a given word can be retrieved under a given viewpoint. However, the time this consumes makes the VB-SDS unsuitable for inference systems. The proposed schemes reduce search space based on the upper bound of a similarity calculation function to increase retrieval speed. An analytical evaluation shows the schemes can achieve a thousand-fold speedup and confirmed through experimental results for a VB-SDS containing about 40,000 words.

A subclass of context-free languages, called pure context-free languages, which is generated by context-free grammar with only one type of symbol (i.e., terminals and nonterminals are not distinguished), is introduced and the problem of identifying from positive data a restricted class of monogenic pure context-free languages (mono-PCF languages, in short) is investigated. The class of mono-PCF languages is incomparable to the class of regular languages. In this paper we show that the class of mono-PCF languages is polynomial time identifiable from positive data. That is, there is an algorithm that, given a mono-PCF language L, identifies from positive data, a grammar generating L, called a monogenic pure context-free grammar (mono-PCF grammar, in short) satisfying the property that the time for updating a conjecture is bounded by O(N3), where N is the sum of lengths of all positive data provided. This is in contrast with another result in this paper that the class of PCF languages is not identifiable in the limit from positive data.

In order to provide a fuzzy system with learning function, numerous studies are being carried out to combine fuzzy systems and neural networks. The self-tuning methods using the descent method have been proposed. The constructive and the destructive methods are more powerful than other methods using neural networks (or descent method). On the other hand the destructive method is superior in the number of rules and inference error and inferior in learning speed to the constructive method. In this paper, we propose a new learning method combining the constructive and the destructive methods. The method is superior in the number of rules, inference error and learning speed to the destructive method. However, it is inferior in learning speed to the constructive method. Therefore, in order to improve learning speed of the proposed method, simplified learning methods are proposed. Some numerical examples are given to show the validity of the proposed methods.

This paper considers properties of language classes with finite elasticity in the viewpoint of set theoretic operations. Finite elasticity was introduced by Wright as a sufficient condition for language classes to be inferable from positive data, and as a property preserved by (not usual) union operation to generate a class of unions of languages. We show that the family of language classes with finite elasticity is closed under not only union but also various operations for language classes such as intersection, concatenation and so on, except complement operation. As a framework defining languages, we introduce restricted elementary formal systems (EFS's for short), called max length-bounded by which any context-sensitive language is definable. We define various operations for EFS's corresponding to usual language operations and also for EFS classes, and investigate closure properties of the family Ge of max length-bounded EFS classes that define classes of languages with finite elasticity. Furthermore, we present theorems characterizing a max length-bounded EFS class in the family Ge, and that for the language class to be inferable from positive data, provided the class is closed under subset operation. From the former, it follows that for any n, a language class definable by max length-bounded EFS's with at most n axioms has finite elasticity. This means that Ge is sufficiently large.

It is shown that a self–recurrent fuzzy inference can cause chaotic responses at least three membership functions, if the inference rules are set to represent nonlinear relations such as pie–kneading transformation. This system has single input and single output both with crisp values, in which membership functions is taken to be triangular. Extensions to infinite memberships are proposed, so as to reproduce the continuum case of one–dimensional logistic map f(x)=Ax(1–x). And bifurcation diagrams are calculated for number N of memberships of 3, 5, 9 and 17. It is found from bifurcation diagrams that different periodic states coexist at the same bifurcation parameter for N9. This indicates multistability necessarily accompanied with hysteresis effects. Therefore, it is concluded that the final states are not uniquely determined by fuzzy inferences with sufficiently large number of memberships.

Nonmonotonic reasoning is a logical inference system which attempts to approximate human commonsense reasoning and is characterized as defeasible: having reasonably drawn a conclusion from some premises we may be forced to retract that conclusion upon learning new facts. This paper introduces a Petri net model for nonmonotonic reasoning with nonmonotonic rules generated by annotated logic programs and the unless operator. In the Petri net model, a fixpoint of a nonmonotonic theory can be represented as a maximal and consistent support of a firing sequence. We propose a structural method for finding extensions (coherent consequences) for a given set of nonmonotonic logic rules. It is based on the T-invariant technique for testing fireability of a goal transition in the Petri net model of Horn clause logic programs.

Among the problems that face ITS designers, the problem of measuring the student knowledge state after concept learning in order to initially adapt a skill acquisition session according to a student's own necessities is a hard one. Typical approaches are the use of some sort of test to assess the student knowledge and choose an initial set of parameters for a session, or use, regardless the particular necessities of a student, a pre-defined set of initial parameters. We consider the fromer to be disrupting for learning and the latter too simple to deal with the broad possibilities that are faced. It is known that students show different behaviors during concept learning depending on the experience, background and actual understanding (the way a student is understanding a concept) during concept learning. Our approach here is to classify the different behaviors through fuzzy proposition and link them with a student model through fuzzy rules to use in an expert system, and with it, select the most suitable problem-solving strategy for each particular student in order to clear his misunderstandings and facilitate the learning of problem-solving skills. The use of probabilistic reasoning (i.e. Bayesian statistics) instead of fuzzy logic is not suitable for the present situation because of the rigidity and precision of the rules that do not allow a proper manipulation of the vagueness involved in the student behavior. We apply this idea to a circuit analysis ITS where the concept learning session is carried out on a Hypertext environment and the skill acquisition session on an interactive problem-solving environment. By tracing the student use of the Hypertext environment we can know the student behavior and use it as a premise in the fuzzy inference.

This paper considers algebraic basic processes, a subset of communicating processes in CCS by Milner, and presents a synthesis algorithm to infer a process that satisfies the properties of the process, represented as fomulae in Hennessy-Milner Logic. The validity of the proposed algorithm can be stated that it synthesizes a process in the limit, which cannot be distinguished from the target one with respect to the strong equivalence.

A fuzzy microprocessor is developed using 1.2 µm CMOS process. The inference scheme for the if-then fuzzy rules consists of three main steps i. e. if-part process, then-part process and defuzzification. In order to realize very high-speed inference and moderate programmability, we introduce three-type different structures i.e. SIMD, logic-in-memory and Wallace tree structures which are suitable for the three main steps. The inference speed including defuzzification is 7.5 MFLIPS which is 12.9 times higher than the previous VLSI implementation, and it can carry out many rules (960 rules) and many input and output variables (16 variables).

This paper is concerned with the problem of (exactly) representing given functions by fuzzy reasoning. We consider function representation by the fuzzy reasoning method using linguistic truth values, which is a generalization of fuzzy reasoning due to Zadeh. Some conditions for functions to be representable are given, by which it is shown that very large class of functions can be representable by this method. Some examples illustrating how to find "if-then rules" for fuzzy reasoning are shown. Further, in the appendix an example is given to show that the generalization is significant for the problem of function representation.

A fuzzy inference processor which performs fuzzy inference with 12-bit resolution input at 200 kFLIPS (Fuzzy Logical Inference Per Second) has been developed. To keep the cost performance, not parallel processing hardware but processor type hardware is employed. Dedicated membership function generators, rule instructions and modified add/divide algorithm are adopted to attain the performance. The membership function generators calculate a membership function value in less than a half clock cycle. Rule instructions calculate the grade of a rule by one instruction. Antecedent processing and consequent processing are pipelined by the modified add/divide algorithm. As a result, total inference time is significantly reduced. For example, in the case of typical inference (about 20 rules with 2 to 4 inputs and 1 output), the total inference needs approximately 100 clock cycles. Furthermore by adding a mechanism to calculate the variance and maximum grade of the final membership function, it is enabled to evaluate the inference reliability. The chip, fabricated by 1 µm CMOS technology, contains 86k transistors in a 7.56.7 mm die size. The chip operates at more than 20 MHz clock frequency at 5 V.

In this paper, we consider the polynomial time inferability from positive data for unions of two tree pattern languages. A tree pattern is a structured pattern known as a term in logic programming, and a tree pattern language is the set of all ground instances of a tree pattern. We present a polynomial time algorithm to find a minimal union of two tree pattern languages containing given examples. Our algorithm can be considered as a natural extension of Plotkin's least generalization algorithm, which finds a minimal single tree pattern language. By using this algorithm, we can realize a consistent and conservative polynomial time inference machine that identifies unions of two tree pattern languages from positive data in the limit.